| *** bradw_ <bradw_!~brad@ipsw-05-b2-v4wan-162355-cust86.vm17.cable.virginm.net> has quit IRC (Ping timeout: 264 seconds) | 00:00 | |
| *** bradw <bradw!~brad@user/bradw> has joined #bzflag | 00:32 | |
| *** Cobra_Fast is now away: vacant | 01:36 | |
| *** Cobra_Fast is back | 01:36 | |
| *** I_Died_Once <I_Died_Once!~I_Died_On@c-73-184-170-223.hsd1.ga.comcast.net> has quit IRC (Ping timeout: 255 seconds) | 01:55 | |
| Agatha | tupone: I skimmed through the updated code, but nothing immediately jumped out at me as being wrong. Could it be a caching issue with the saved map; the texture gets serialized to disk in a funky way . . . | 02:50 |
|---|---|---|
| Agatha | Also, the project versions for MSVC seem to be VS 2017. It's difficult to even *download* VS 2017 anymore. The build seems to work fine with VS 2022. Let's upgrade, hey? | 02:52 |
| blast007 | Agatha: no, it's definitely something with the code change. Simply compiling from the commit prior to that commit works fine. | 02:52 |
| blast007 | it's also not difficult to download 2017 https://visualstudio.microsoft.com/thank-you-downloading-visual-studio/?sku=Community&rel=15 | 02:53 |
| Agatha | That's the page that shows . . . if you download VS 2022. I poked around a bit; the older versions aren't there. If you really want to download them, you have to search for it specifically, to get the deliberately orphaned page https://visualstudio.microsoft.com/vs/older-downloads/ | 02:55 |
| Agatha | It's worth noting that VS 2022 supports the VS 2017 build tools seamlessly, so this amounts to an editor upgrade at worst. | 02:55 |
| blast007 | is there something in 2022 that we need? | 02:58 |
| Agatha | By the same token as it working in both, technically no . . . albeit, one expects the VS 2022 build to be faster and more standards-compliant, in a better editor that they don't have to take the extra step to retarget builds to . . . | 03:01 |
| blast007 | also doesn't support Windows 7 | 03:04 |
| blast007 | and can't target Vista | 03:06 |
| blast007 | we still have at least one or two players still running on XP, though that will break once I enforce HTTPS | 03:07 |
| blast007 | (since the last XP build was 2.4.2 I think?) | 03:07 |
| *** Sgeo_ <Sgeo_!~Sgeo@user/sgeo> has joined #bzflag | 03:26 | |
| *** Sgeo <Sgeo!~Sgeo@user/sgeo> has quit IRC (Ping timeout: 268 seconds) | 03:27 | |
| Agatha | @tupone, @blast007: aight I think I know the problem with the textures. TL;DR add `glPixelStorei(GL_UNPACK_ALIGNMENT,1);` somewhere (I suggest you add it, along with the `glPixelStorei(GL_PACK_ALIGNMENT,1);` in "clientCommands.cxx", to some sort of global OpenGL initialization). | 03:48 |
| Agatha | The reason you need this call is pretty simple: the default is 4, meaning GL expects a multiple of 4 bytes per row. Your new texture code compacts the memory from RGBA to RGB / L / L+A if possible, which can result in a different alignment. In this case, the texture is 739 pixels across. You new code compacts this from 739*4=2956 bytes RGBA to 739*3=2217 bytes RGB, which is not a multiple of 4. OpenGL rounds up to the next multiple of 4, skipping what it | 03:48 |
| Agatha | assumes are intentional padding bytes, before starting on the next line. In this case, that's three bytes. However, those three bytes are not padding; they *should* have been the next pixel. These are skipped, so OpenGL is effectively skipping one pixel. One pixel is skipped on every row, so you get a 45 degree angle. The thing doesn't segfault because the original memory store never got shrunk. This problem didn't happen with the previous code because the | 03:48 |
| Agatha | previous code used RGBA for the upload, which is a multiple of 4. | 03:48 |
| SpringTank | oh so that's why the textures on some maps are shifted! | 03:54 |
| *** Sgeo__ <Sgeo__!~Sgeo@user/sgeo> has joined #bzflag | 04:16 | |
| *** Sgeo_ <Sgeo_!~Sgeo@user/sgeo> has quit IRC (Ping timeout: 255 seconds) | 04:19 | |
| Agatha | I've noticed this general category of bug for a long time, even reported one. I long suspected it was this kind of issue; this should fix it for good. | 04:44 |
| Agatha | (Here's that issue I alluded to, BTW: https://github.com/BZFlag-Dev/bzflag/issues/258 I wouldn't be surprised if this is caused by a similar problem.) | 06:08 |
| *** Sgeo__ <Sgeo__!~Sgeo@user/sgeo> has quit IRC (Read error: Connection reset by peer) | 07:09 | |
| *** FastLizard4 is back | 08:09 | |
| *** blast007[m] <blast007[m]!~blast007m@2001:470:69fc:105::7ec> has quit IRC (Quit: You have been kicked for being idle) | 08:52 | |
| *** I_Died_Once <I_Died_Once!~I_Died_On@c-73-184-170-223.hsd1.ga.comcast.net> has joined #bzflag | 09:01 | |
| *** FastLizard4 is now away: AWAY from keyboard | 09:01 | |
| *** FastLizard4 is now away: GONE - Screen Detached and Disconnected from IRC (I'm probably asleep, at work, or doing something in real life) | 09:27 | |
| blast007 | Agatha: in the function that takes a screenshot? | 10:17 |
| *** blast007[m] <blast007[m]!~blast007m@2001:470:69fc:105::7ec> has joined #bzflag | 10:17 | |
| macsforme | great analysis, Agatha... thank you | 13:51 |
| Agatha | @blast007 yes. An analogous potential issue exists reading data *from* OpenGL as well (e.g. taking a screenshot), which one solves by setting the *pack* alignment. | 14:29 |
| Agatha | Anywho, the parameters are global GL state, so I'd put them both into some initializer somewhere. | 14:30 |
| *** FastLizard4 is back | 18:36 | |
| *** Cobra_Fast is now away: vacant | 19:03 | |
| *** Cobra_Fast is back | 19:03 | |
| *** FastLizard4 is now away: AWAY from keyboard | 19:06 | |
| SpringTank | have you been able to test this and recompile? | 19:25 |
| tupone | I did | 19:34 |
| Agatha | Yeah I put a `glPixelStorei(GL_UNPACK_ALIGNMENT,1);` somewhere and confirmed it fixes the problem, at least on that map. | 19:34 |
| BZNotify | 2.4 @ bzflag: atupone pushed 1 commit (https://github.com/BZFlag-Dev/bzflag/compare/ec6fff321415...1dcf6ebf029e): | 19:35 |
| BZNotify | 2.4 @ bzflag: atupone 1dcf6e: Fix packing of texture, thanks to Agatha (https://github.com/BZFlag-Dev/bzflag/commit/1dcf6ebf029ee095685e0f6c183ae0a4ad0f0117) | 19:35 |
| Agatha | Nice :) Note the other pack alignment setting in "clientCommands.cxx" is now redundant, and so should be removed. | 19:37 |
| SpringTank | why I never learned openGL https://i.redd.it/gyufhogm52l91.jpg | 19:54 |
| BZNotify | 2.4 @ bzflag: atupone pushed 1 commit (https://github.com/BZFlag-Dev/bzflag/compare/1dcf6ebf029e...04788c541f72): | 20:03 |
| BZNotify | 2.4 @ bzflag: atupone 04788c: Remove pack alignment, global now (https://github.com/BZFlag-Dev/bzflag/commit/04788c541f72882d3083e26c23434bb42e3ee384) | 20:03 |
| *** Sgeo <Sgeo!~Sgeo@user/sgeo> has joined #bzflag | 22:33 | |
| *** yuitimothy_ is back | 23:02 | |
| *** Cobra_Fast is now away: vacant | 23:28 | |
| *** Cobra_Fast is back | 23:28 | |
Generated by irclog2html.py 2.17.3 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!
